2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "config-host.h"
25 #include "qemu-common.h"
27 #include "block_int.h"
29 #include "qemu-objects.h"
32 #include <sys/types.h>
34 #include <sys/ioctl.h>
35 #include <sys/queue.h>
45 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
46 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
47 BlockDriverCompletionFunc
*cb
, void *opaque
);
48 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
49 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
50 BlockDriverCompletionFunc
*cb
, void *opaque
);
51 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
52 BlockDriverCompletionFunc
*cb
, void *opaque
);
53 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
54 uint8_t *buf
, int nb_sectors
);
55 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
56 const uint8_t *buf
, int nb_sectors
);
57 static BlockDriver
*find_protocol(const char *filename
);
59 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
60 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
62 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
63 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
65 /* If non-zero, use only whitelisted block drivers */
66 static int use_bdrv_whitelist
;
68 int path_is_absolute(const char *path
)
72 /* specific case for names like: "\\.\d:" */
73 if (*path
== '/' || *path
== '\\')
76 p
= strchr(path
, ':');
82 return (*p
== '/' || *p
== '\\');
88 /* if filename is absolute, just copy it to dest. Otherwise, build a
89 path to it by considering it is relative to base_path. URL are
91 void path_combine(char *dest
, int dest_size
,
92 const char *base_path
,
100 if (path_is_absolute(filename
)) {
101 pstrcpy(dest
, dest_size
, filename
);
103 p
= strchr(base_path
, ':');
108 p1
= strrchr(base_path
, '/');
112 p2
= strrchr(base_path
, '\\');
124 if (len
> dest_size
- 1)
126 memcpy(dest
, base_path
, len
);
128 pstrcat(dest
, dest_size
, filename
);
132 void bdrv_register(BlockDriver
*bdrv
)
134 if (!bdrv
->bdrv_aio_readv
) {
135 /* add AIO emulation layer */
136 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
137 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
138 } else if (!bdrv
->bdrv_read
) {
139 /* add synchronous IO emulation layer */
140 bdrv
->bdrv_read
= bdrv_read_em
;
141 bdrv
->bdrv_write
= bdrv_write_em
;
144 if (!bdrv
->bdrv_aio_flush
)
145 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
147 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
150 /* create a new block device (by default it is empty) */
151 BlockDriverState
*bdrv_new(const char *device_name
)
153 BlockDriverState
*bs
;
155 bs
= qemu_mallocz(sizeof(BlockDriverState
));
156 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
157 if (device_name
[0] != '\0') {
158 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
163 BlockDriver
*bdrv_find_format(const char *format_name
)
166 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
167 if (!strcmp(drv1
->format_name
, format_name
)) {
174 static int bdrv_is_whitelisted(BlockDriver
*drv
)
176 static const char *whitelist
[] = {
177 CONFIG_BDRV_WHITELIST
182 return 1; /* no whitelist, anything goes */
184 for (p
= whitelist
; *p
; p
++) {
185 if (!strcmp(drv
->format_name
, *p
)) {
192 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
194 BlockDriver
*drv
= bdrv_find_format(format_name
);
195 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
198 int bdrv_create(BlockDriver
*drv
, const char* filename
,
199 QEMUOptionParameter
*options
)
201 if (!drv
->bdrv_create
)
204 return drv
->bdrv_create(filename
, options
);
207 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
211 drv
= find_protocol(filename
);
213 drv
= bdrv_find_format("file");
216 return bdrv_create(drv
, filename
, options
);
220 void get_tmp_filename(char *filename
, int size
)
222 char temp_dir
[MAX_PATH
];
224 GetTempPath(MAX_PATH
, temp_dir
);
225 GetTempFileName(temp_dir
, "qem", 0, filename
);
228 void get_tmp_filename(char *filename
, int size
)
232 /* XXX: race condition possible */
233 tmpdir
= getenv("TMPDIR");
236 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
237 fd
= mkstemp(filename
);
243 static int is_windows_drive_prefix(const char *filename
)
245 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
246 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
250 int is_windows_drive(const char *filename
)
252 if (is_windows_drive_prefix(filename
) &&
255 if (strstart(filename
, "\\\\.\\", NULL
) ||
256 strstart(filename
, "//./", NULL
))
263 * Detect host devices. By convention, /dev/cdrom[N] is always
264 * recognized as a host CDROM.
266 static BlockDriver
*find_hdev_driver(const char *filename
)
268 int score_max
= 0, score
;
269 BlockDriver
*drv
= NULL
, *d
;
271 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
272 if (d
->bdrv_probe_device
) {
273 score
= d
->bdrv_probe_device(filename
);
274 if (score
> score_max
) {
284 static BlockDriver
*find_protocol(const char *filename
)
292 /* TODO Drivers without bdrv_file_open must be specified explicitly */
295 is_drive
= is_windows_drive(filename
) ||
296 is_windows_drive_prefix(filename
);
300 p
= strchr(filename
, ':');
301 if (!p
|| is_drive
) {
302 drv1
= find_hdev_driver(filename
);
304 drv1
= bdrv_find_format("file");
309 if (len
> sizeof(protocol
) - 1)
310 len
= sizeof(protocol
) - 1;
311 memcpy(protocol
, filename
, len
);
312 protocol
[len
] = '\0';
313 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
314 if (drv1
->protocol_name
&&
315 !strcmp(drv1
->protocol_name
, protocol
)) {
322 static BlockDriver
*find_image_format(const char *filename
)
324 int ret
, score
, score_max
;
325 BlockDriver
*drv1
, *drv
;
327 BlockDriverState
*bs
;
329 drv
= find_protocol(filename
);
330 /* no need to test disk image formats for vvfat */
331 if (drv
&& strcmp(drv
->format_name
, "vvfat") == 0)
334 ret
= bdrv_file_open(&bs
, filename
, 0);
337 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
345 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
346 if (drv1
->bdrv_probe
) {
347 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
348 if (score
> score_max
) {
358 * Set the current 'total_sectors' value
360 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
362 BlockDriver
*drv
= bs
->drv
;
364 /* query actual device if possible, otherwise just trust the hint */
365 if (drv
->bdrv_getlength
) {
366 int64_t length
= drv
->bdrv_getlength(bs
);
370 hint
= length
>> BDRV_SECTOR_BITS
;
373 bs
->total_sectors
= hint
;
378 * Common part for opening disk images and files
380 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
381 int flags
, BlockDriver
*drv
)
388 bs
->total_sectors
= 0;
389 bs
->is_temporary
= 0;
392 bs
->open_flags
= flags
;
393 /* buffer_alignment defaulted to 512, drivers can change this value */
394 bs
->buffer_alignment
= 512;
396 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
398 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
403 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
406 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
407 * write cache to the guest. We do need the fdatasync to flush
408 * out transactions for block allocations, and we maybe have a
409 * volatile write cache in our backing device to deal with.
411 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
412 bs
->enable_write_cache
= 1;
415 * Clear flags that are internal to the block layer before opening the
418 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
421 * Snapshots should be writeable.
423 if (bs
->is_temporary
) {
424 open_flags
|= BDRV_O_RDWR
;
427 /* Open the image, either directly or using a protocol */
428 if (drv
->bdrv_file_open
) {
429 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
431 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
433 ret
= drv
->bdrv_open(bs
, open_flags
);
441 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
443 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
449 if (bs
->is_temporary
) {
457 bdrv_delete(bs
->file
);
460 qemu_free(bs
->opaque
);
467 * Opens a file using a protocol (file, host_device, nbd, ...)
469 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
471 BlockDriverState
*bs
;
475 drv
= find_protocol(filename
);
481 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
492 * Opens a disk image (raw, qcow2, vmdk, ...)
494 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
499 if (flags
& BDRV_O_SNAPSHOT
) {
500 BlockDriverState
*bs1
;
503 BlockDriver
*bdrv_qcow2
;
504 QEMUOptionParameter
*options
;
505 char tmp_filename
[PATH_MAX
];
506 char backing_filename
[PATH_MAX
];
508 /* if snapshot, we create a temporary backing file and open it
509 instead of opening 'filename' directly */
511 /* if there is a backing file, use it */
513 ret
= bdrv_open(bs1
, filename
, 0, drv
);
518 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
520 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
525 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
527 /* Real path is meaningless for protocols */
529 snprintf(backing_filename
, sizeof(backing_filename
),
531 else if (!realpath(filename
, backing_filename
))
534 bdrv_qcow2
= bdrv_find_format("qcow2");
535 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
537 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
538 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
540 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
544 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
545 free_option_parameters(options
);
550 filename
= tmp_filename
;
552 bs
->is_temporary
= 1;
555 /* Find the right image format driver */
557 drv
= find_image_format(filename
);
562 goto unlink_and_fail
;
566 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
568 goto unlink_and_fail
;
571 /* If there is a backing file, use it */
572 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
573 char backing_filename
[PATH_MAX
];
575 BlockDriver
*back_drv
= NULL
;
577 bs
->backing_hd
= bdrv_new("");
578 path_combine(backing_filename
, sizeof(backing_filename
),
579 filename
, bs
->backing_file
);
580 if (bs
->backing_format
[0] != '\0')
581 back_drv
= bdrv_find_format(bs
->backing_format
);
583 /* backing files always opened read-only */
585 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
587 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
592 if (bs
->is_temporary
) {
593 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
595 /* base image inherits from "parent" */
596 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
600 if (!bdrv_key_required(bs
)) {
601 /* call the change callback */
602 bs
->media_changed
= 1;
604 bs
->change_cb(bs
->change_opaque
);
610 if (bs
->is_temporary
) {
616 void bdrv_close(BlockDriverState
*bs
)
619 if (bs
->backing_hd
) {
620 bdrv_delete(bs
->backing_hd
);
621 bs
->backing_hd
= NULL
;
623 bs
->drv
->bdrv_close(bs
);
624 qemu_free(bs
->opaque
);
626 if (bs
->is_temporary
) {
627 unlink(bs
->filename
);
633 if (bs
->file
!= NULL
) {
634 bdrv_close(bs
->file
);
637 /* call the change callback */
638 bs
->media_changed
= 1;
640 bs
->change_cb(bs
->change_opaque
);
644 void bdrv_delete(BlockDriverState
*bs
)
646 /* remove from list, if necessary */
647 if (bs
->device_name
[0] != '\0') {
648 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
652 if (bs
->file
!= NULL
) {
653 bdrv_delete(bs
->file
);
660 * Run consistency checks on an image
662 * Returns the number of errors or -errno when an internal error occurs
664 int bdrv_check(BlockDriverState
*bs
)
666 if (bs
->drv
->bdrv_check
== NULL
) {
670 return bs
->drv
->bdrv_check(bs
);
673 /* commit COW file into the raw image */
674 int bdrv_commit(BlockDriverState
*bs
)
676 BlockDriver
*drv
= bs
->drv
;
677 int64_t i
, total_sectors
;
678 int n
, j
, ro
, open_flags
;
679 int ret
= 0, rw_ret
= 0;
680 unsigned char sector
[512];
682 BlockDriverState
*bs_rw
, *bs_ro
;
687 if (!bs
->backing_hd
) {
691 if (bs
->backing_hd
->keep_read_only
) {
695 ro
= bs
->backing_hd
->read_only
;
696 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
697 open_flags
= bs
->backing_hd
->open_flags
;
701 bdrv_delete(bs
->backing_hd
);
702 bs
->backing_hd
= NULL
;
703 bs_rw
= bdrv_new("");
704 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, drv
);
707 /* try to re-open read-only */
708 bs_ro
= bdrv_new("");
709 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, drv
);
712 /* drive not functional anymore */
716 bs
->backing_hd
= bs_ro
;
719 bs
->backing_hd
= bs_rw
;
722 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
723 for (i
= 0; i
< total_sectors
;) {
724 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
725 for(j
= 0; j
< n
; j
++) {
726 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
731 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
742 if (drv
->bdrv_make_empty
) {
743 ret
= drv
->bdrv_make_empty(bs
);
748 * Make sure all data we wrote to the backing device is actually
752 bdrv_flush(bs
->backing_hd
);
758 bdrv_delete(bs
->backing_hd
);
759 bs
->backing_hd
= NULL
;
760 bs_ro
= bdrv_new("");
761 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, drv
);
764 /* drive not functional anymore */
768 bs
->backing_hd
= bs_ro
;
769 bs
->backing_hd
->keep_read_only
= 0;
778 * -EINVAL - backing format specified, but no file
779 * -ENOSPC - can't update the backing file because no space is left in the
781 * -ENOTSUP - format driver doesn't support changing the backing file
783 int bdrv_change_backing_file(BlockDriverState
*bs
,
784 const char *backing_file
, const char *backing_fmt
)
786 BlockDriver
*drv
= bs
->drv
;
788 if (drv
->bdrv_change_backing_file
!= NULL
) {
789 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
795 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
800 if (!bdrv_is_inserted(bs
))
806 len
= bdrv_getlength(bs
);
811 if ((offset
> len
) || (len
- offset
< size
))
817 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
820 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
823 /* return < 0 if error. See bdrv_write() for the return codes */
824 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
825 uint8_t *buf
, int nb_sectors
)
827 BlockDriver
*drv
= bs
->drv
;
831 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
834 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
837 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
838 int nb_sectors
, int dirty
)
841 unsigned long val
, idx
, bit
;
843 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
844 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
846 for (; start
<= end
; start
++) {
847 idx
= start
/ (sizeof(unsigned long) * 8);
848 bit
= start
% (sizeof(unsigned long) * 8);
849 val
= bs
->dirty_bitmap
[idx
];
851 if (!(val
& (1 << bit
))) {
856 if (val
& (1 << bit
)) {
861 bs
->dirty_bitmap
[idx
] = val
;
865 /* Return < 0 if error. Important errors are:
866 -EIO generic I/O error (may happen for all errors)
867 -ENOMEDIUM No media inserted.
868 -EINVAL Invalid sector number or nb_sectors
869 -EACCES Trying to write a read-only device
871 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
872 const uint8_t *buf
, int nb_sectors
)
874 BlockDriver
*drv
= bs
->drv
;
879 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
882 if (bs
->dirty_bitmap
) {
883 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
886 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
887 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
890 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
893 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
894 void *buf
, int count1
)
896 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
897 int len
, nb_sectors
, count
;
902 /* first read to align to sector start */
903 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
906 sector_num
= offset
>> BDRV_SECTOR_BITS
;
908 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
910 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
918 /* read the sectors "in place" */
919 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
920 if (nb_sectors
> 0) {
921 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
923 sector_num
+= nb_sectors
;
924 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
929 /* add data from the last sector */
931 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
933 memcpy(buf
, tmp_buf
, count
);
938 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
939 const void *buf
, int count1
)
941 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
942 int len
, nb_sectors
, count
;
947 /* first write to align to sector start */
948 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
951 sector_num
= offset
>> BDRV_SECTOR_BITS
;
953 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
955 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
956 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
965 /* write the sectors "in place" */
966 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
967 if (nb_sectors
> 0) {
968 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
970 sector_num
+= nb_sectors
;
971 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
976 /* add data from the last sector */
978 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
980 memcpy(tmp_buf
, buf
, count
);
981 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
988 * Truncate file to 'offset' bytes (needed only for file protocols)
990 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
992 BlockDriver
*drv
= bs
->drv
;
996 if (!drv
->bdrv_truncate
)
1000 ret
= drv
->bdrv_truncate(bs
, offset
);
1002 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1008 * Length of a file in bytes. Return < 0 if error or unknown.
1010 int64_t bdrv_getlength(BlockDriverState
*bs
)
1012 BlockDriver
*drv
= bs
->drv
;
1016 /* Fixed size devices use the total_sectors value for speed instead of
1017 issuing a length query (like lseek) on each call. Also, legacy block
1018 drivers don't provide a bdrv_getlength function and must use
1020 if (!bs
->growable
|| !drv
->bdrv_getlength
) {
1021 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1023 return drv
->bdrv_getlength(bs
);
1026 /* return 0 as number of sectors if no device present or error */
1027 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1030 length
= bdrv_getlength(bs
);
1034 length
= length
>> BDRV_SECTOR_BITS
;
1035 *nb_sectors_ptr
= length
;
1039 uint8_t boot_ind
; /* 0x80 - active */
1040 uint8_t head
; /* starting head */
1041 uint8_t sector
; /* starting sector */
1042 uint8_t cyl
; /* starting cylinder */
1043 uint8_t sys_ind
; /* What partition type */
1044 uint8_t end_head
; /* end head */
1045 uint8_t end_sector
; /* end sector */
1046 uint8_t end_cyl
; /* end cylinder */
1047 uint32_t start_sect
; /* starting sector counting from 0 */
1048 uint32_t nr_sects
; /* nr of sectors in partition */
1049 } __attribute__((packed
));
1051 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1052 static int guess_disk_lchs(BlockDriverState
*bs
,
1053 int *pcylinders
, int *pheads
, int *psectors
)
1056 int ret
, i
, heads
, sectors
, cylinders
;
1057 struct partition
*p
;
1059 uint64_t nb_sectors
;
1061 bdrv_get_geometry(bs
, &nb_sectors
);
1063 ret
= bdrv_read(bs
, 0, buf
, 1);
1066 /* test msdos magic */
1067 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1069 for(i
= 0; i
< 4; i
++) {
1070 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1071 nr_sects
= le32_to_cpu(p
->nr_sects
);
1072 if (nr_sects
&& p
->end_head
) {
1073 /* We make the assumption that the partition terminates on
1074 a cylinder boundary */
1075 heads
= p
->end_head
+ 1;
1076 sectors
= p
->end_sector
& 63;
1079 cylinders
= nb_sectors
/ (heads
* sectors
);
1080 if (cylinders
< 1 || cylinders
> 16383)
1083 *psectors
= sectors
;
1084 *pcylinders
= cylinders
;
1086 printf("guessed geometry: LCHS=%d %d %d\n",
1087 cylinders
, heads
, sectors
);
1095 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1097 int translation
, lba_detected
= 0;
1098 int cylinders
, heads
, secs
;
1099 uint64_t nb_sectors
;
1101 /* if a geometry hint is available, use it */
1102 bdrv_get_geometry(bs
, &nb_sectors
);
1103 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1104 translation
= bdrv_get_translation_hint(bs
);
1105 if (cylinders
!= 0) {
1110 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1112 /* if heads > 16, it means that a BIOS LBA
1113 translation was active, so the default
1114 hardware geometry is OK */
1116 goto default_geometry
;
1121 /* disable any translation to be in sync with
1122 the logical geometry */
1123 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1124 bdrv_set_translation_hint(bs
,
1125 BIOS_ATA_TRANSLATION_NONE
);
1130 /* if no geometry, use a standard physical disk geometry */
1131 cylinders
= nb_sectors
/ (16 * 63);
1133 if (cylinders
> 16383)
1135 else if (cylinders
< 2)
1140 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1141 if ((*pcyls
* *pheads
) <= 131072) {
1142 bdrv_set_translation_hint(bs
,
1143 BIOS_ATA_TRANSLATION_LARGE
);
1145 bdrv_set_translation_hint(bs
,
1146 BIOS_ATA_TRANSLATION_LBA
);
1150 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1154 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1155 int cyls
, int heads
, int secs
)
1162 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1165 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1166 type
== BDRV_TYPE_FLOPPY
));
1169 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1171 bs
->translation
= translation
;
1174 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1175 int *pcyls
, int *pheads
, int *psecs
)
1178 *pheads
= bs
->heads
;
1182 int bdrv_get_type_hint(BlockDriverState
*bs
)
1187 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1189 return bs
->translation
;
1192 int bdrv_is_removable(BlockDriverState
*bs
)
1194 return bs
->removable
;
1197 int bdrv_is_read_only(BlockDriverState
*bs
)
1199 return bs
->read_only
;
1202 int bdrv_is_sg(BlockDriverState
*bs
)
1207 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1209 return bs
->enable_write_cache
;
1212 /* XXX: no longer used */
1213 void bdrv_set_change_cb(BlockDriverState
*bs
,
1214 void (*change_cb
)(void *opaque
), void *opaque
)
1216 bs
->change_cb
= change_cb
;
1217 bs
->change_opaque
= opaque
;
1220 int bdrv_is_encrypted(BlockDriverState
*bs
)
1222 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1224 return bs
->encrypted
;
1227 int bdrv_key_required(BlockDriverState
*bs
)
1229 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1231 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1233 return (bs
->encrypted
&& !bs
->valid_key
);
1236 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1239 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1240 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1246 if (!bs
->encrypted
) {
1248 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1251 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1254 } else if (!bs
->valid_key
) {
1256 /* call the change callback now, we skipped it on open */
1257 bs
->media_changed
= 1;
1259 bs
->change_cb(bs
->change_opaque
);
1264 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1269 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1273 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1278 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1279 it(opaque
, drv
->format_name
);
1283 BlockDriverState
*bdrv_find(const char *name
)
1285 BlockDriverState
*bs
;
1287 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1288 if (!strcmp(name
, bs
->device_name
)) {
1295 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1297 BlockDriverState
*bs
;
1299 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1304 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1306 return bs
->device_name
;
1309 void bdrv_flush(BlockDriverState
*bs
)
1311 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1312 bs
->drv
->bdrv_flush(bs
);
1315 void bdrv_flush_all(void)
1317 BlockDriverState
*bs
;
1319 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1320 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1321 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1327 int bdrv_has_zero_init(BlockDriverState
*bs
)
1331 if (bs
->drv
->no_zero_init
) {
1333 } else if (bs
->file
) {
1334 return bdrv_has_zero_init(bs
->file
);
1341 * Returns true iff the specified sector is present in the disk image. Drivers
1342 * not implementing the functionality are assumed to not support backing files,
1343 * hence all their sectors are reported as allocated.
1345 * 'pnum' is set to the number of sectors (including and immediately following
1346 * the specified sector) that are known to be in the same
1347 * allocated/unallocated state.
1349 * 'nb_sectors' is the max value 'pnum' should be set to.
1351 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1355 if (!bs
->drv
->bdrv_is_allocated
) {
1356 if (sector_num
>= bs
->total_sectors
) {
1360 n
= bs
->total_sectors
- sector_num
;
1361 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1364 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1367 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1368 BlockMonEventAction action
, int is_read
)
1371 const char *action_str
;
1374 case BDRV_ACTION_REPORT
:
1375 action_str
= "report";
1377 case BDRV_ACTION_IGNORE
:
1378 action_str
= "ignore";
1380 case BDRV_ACTION_STOP
:
1381 action_str
= "stop";
1387 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1390 is_read
? "read" : "write");
1391 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1393 qobject_decref(data
);
1396 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1399 Monitor
*mon
= opaque
;
1401 bs_dict
= qobject_to_qdict(obj
);
1403 monitor_printf(mon
, "%s: type=%s removable=%d",
1404 qdict_get_str(bs_dict
, "device"),
1405 qdict_get_str(bs_dict
, "type"),
1406 qdict_get_bool(bs_dict
, "removable"));
1408 if (qdict_get_bool(bs_dict
, "removable")) {
1409 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1412 if (qdict_haskey(bs_dict
, "inserted")) {
1413 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1415 monitor_printf(mon
, " file=");
1416 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1417 if (qdict_haskey(qdict
, "backing_file")) {
1418 monitor_printf(mon
, " backing_file=");
1419 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1421 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1422 qdict_get_bool(qdict
, "ro"),
1423 qdict_get_str(qdict
, "drv"),
1424 qdict_get_bool(qdict
, "encrypted"));
1426 monitor_printf(mon
, " [not inserted]");
1429 monitor_printf(mon
, "\n");
1432 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1434 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1438 * bdrv_info(): Block devices information
1440 * Each block device information is stored in a QDict and the
1441 * returned QObject is a QList of all devices.
1443 * The QDict contains the following:
1445 * - "device": device name
1446 * - "type": device type
1447 * - "removable": true if the device is removable, false otherwise
1448 * - "locked": true if the device is locked, false otherwise
1449 * - "inserted": only present if the device is inserted, it is a QDict
1450 * containing the following:
1451 * - "file": device file name
1452 * - "ro": true if read-only, false otherwise
1453 * - "drv": driver format name
1454 * - "backing_file": backing file name if one is used
1455 * - "encrypted": true if encrypted, false otherwise
1459 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1460 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1461 * { "device": "floppy0", "type": "floppy", "removable": true,
1462 * "locked": false } ]
1464 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1467 BlockDriverState
*bs
;
1469 bs_list
= qlist_new();
1471 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1473 const char *type
= "unknown";
1479 case BDRV_TYPE_CDROM
:
1482 case BDRV_TYPE_FLOPPY
:
1487 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1488 "'removable': %i, 'locked': %i }",
1489 bs
->device_name
, type
, bs
->removable
,
1494 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1496 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1497 "'encrypted': %i }",
1498 bs
->filename
, bs
->read_only
,
1499 bs
->drv
->format_name
,
1500 bdrv_is_encrypted(bs
));
1501 if (bs
->backing_file
[0] != '\0') {
1502 QDict
*qdict
= qobject_to_qdict(obj
);
1503 qdict_put(qdict
, "backing_file",
1504 qstring_from_str(bs
->backing_file
));
1507 qdict_put_obj(bs_dict
, "inserted", obj
);
1509 qlist_append_obj(bs_list
, bs_obj
);
1512 *ret_data
= QOBJECT(bs_list
);
1515 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1518 Monitor
*mon
= opaque
;
1520 qdict
= qobject_to_qdict(data
);
1521 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1523 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1524 monitor_printf(mon
, " rd_bytes=%" PRId64
1525 " wr_bytes=%" PRId64
1526 " rd_operations=%" PRId64
1527 " wr_operations=%" PRId64
1529 qdict_get_int(qdict
, "rd_bytes"),
1530 qdict_get_int(qdict
, "wr_bytes"),
1531 qdict_get_int(qdict
, "rd_operations"),
1532 qdict_get_int(qdict
, "wr_operations"));
1535 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1537 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1540 static QObject
* bdrv_info_stats_bs(BlockDriverState
*bs
)
1545 res
= qobject_from_jsonf("{ 'stats': {"
1546 "'rd_bytes': %" PRId64
","
1547 "'wr_bytes': %" PRId64
","
1548 "'rd_operations': %" PRId64
","
1549 "'wr_operations': %" PRId64
","
1550 "'wr_highest_offset': %" PRId64
1552 bs
->rd_bytes
, bs
->wr_bytes
,
1553 bs
->rd_ops
, bs
->wr_ops
,
1554 bs
->wr_highest_sector
* 512);
1555 dict
= qobject_to_qdict(res
);
1557 if (*bs
->device_name
) {
1558 qdict_put(dict
, "device", qstring_from_str(bs
->device_name
));
1562 QObject
*parent
= bdrv_info_stats_bs(bs
->file
);
1563 qdict_put_obj(dict
, "parent", parent
);
1570 * bdrv_info_stats(): show block device statistics
1572 * Each device statistic information is stored in a QDict and
1573 * the returned QObject is a QList of all devices.
1575 * The QDict contains the following:
1577 * - "device": device name
1578 * - "stats": A QDict with the statistics information, it contains:
1579 * - "rd_bytes": bytes read
1580 * - "wr_bytes": bytes written
1581 * - "rd_operations": read operations
1582 * - "wr_operations": write operations
1583 * - "wr_highest_offset": Highest offset of a sector written since the
1584 * BlockDriverState has been opened
1585 * - "parent": A QDict recursively holding the statistics of the underlying
1586 * protocol (e.g. the host file for a qcow2 image). If there is no
1587 * underlying protocol, this field is omitted.
1591 * [ { "device": "ide0-hd0",
1592 * "stats": { "rd_bytes": 512,
1594 * "rd_operations": 1,
1595 * "wr_operations": 0,
1596 * "wr_highest_offset": 0 },
1598 * "stats": { "rd_bytes": 1024,
1600 * "rd_operations": 2,
1601 * "wr_operations": 0,
1602 * "wr_highest_offset": 0,
1604 * { "device": "ide1-cd0",
1605 * "stats": { "rd_bytes": 0,
1607 * "rd_operations": 0,
1608 * "wr_operations": 0,
1609 * "wr_highest_offset": 0 } },
1611 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1615 BlockDriverState
*bs
;
1617 devices
= qlist_new();
1619 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1620 obj
= bdrv_info_stats_bs(bs
);
1621 qlist_append_obj(devices
, obj
);
1624 *ret_data
= QOBJECT(devices
);
1627 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1629 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1630 return bs
->backing_file
;
1631 else if (bs
->encrypted
)
1632 return bs
->filename
;
1637 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1638 char *filename
, int filename_size
)
1640 if (!bs
->backing_file
) {
1641 pstrcpy(filename
, filename_size
, "");
1643 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1647 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1648 const uint8_t *buf
, int nb_sectors
)
1650 BlockDriver
*drv
= bs
->drv
;
1653 if (!drv
->bdrv_write_compressed
)
1655 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1658 if (bs
->dirty_bitmap
) {
1659 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1662 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1665 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1667 BlockDriver
*drv
= bs
->drv
;
1670 if (!drv
->bdrv_get_info
)
1672 memset(bdi
, 0, sizeof(*bdi
));
1673 return drv
->bdrv_get_info(bs
, bdi
);
1676 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1677 int64_t pos
, int size
)
1679 BlockDriver
*drv
= bs
->drv
;
1682 if (!drv
->bdrv_save_vmstate
)
1684 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1687 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1688 int64_t pos
, int size
)
1690 BlockDriver
*drv
= bs
->drv
;
1693 if (!drv
->bdrv_load_vmstate
)
1695 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1698 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1700 BlockDriver
*drv
= bs
->drv
;
1702 if (!drv
|| !drv
->bdrv_debug_event
) {
1706 return drv
->bdrv_debug_event(bs
, event
);
1710 /**************************************************************/
1711 /* handling of snapshots */
1713 int bdrv_snapshot_create(BlockDriverState
*bs
,
1714 QEMUSnapshotInfo
*sn_info
)
1716 BlockDriver
*drv
= bs
->drv
;
1719 if (!drv
->bdrv_snapshot_create
)
1721 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1724 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1725 const char *snapshot_id
)
1727 BlockDriver
*drv
= bs
->drv
;
1730 if (!drv
->bdrv_snapshot_goto
)
1732 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1735 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1737 BlockDriver
*drv
= bs
->drv
;
1740 if (!drv
->bdrv_snapshot_delete
)
1742 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1745 int bdrv_snapshot_list(BlockDriverState
*bs
,
1746 QEMUSnapshotInfo
**psn_info
)
1748 BlockDriver
*drv
= bs
->drv
;
1751 if (!drv
->bdrv_snapshot_list
)
1753 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1756 #define NB_SUFFIXES 4
1758 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1760 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1765 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1768 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1769 if (size
< (10 * base
)) {
1770 snprintf(buf
, buf_size
, "%0.1f%c",
1771 (double)size
/ base
,
1774 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1775 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1776 ((size
+ (base
>> 1)) / base
),
1786 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1788 char buf1
[128], date_buf
[128], clock_buf
[128];
1798 snprintf(buf
, buf_size
,
1799 "%-10s%-20s%7s%20s%15s",
1800 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1804 ptm
= localtime(&ti
);
1805 strftime(date_buf
, sizeof(date_buf
),
1806 "%Y-%m-%d %H:%M:%S", ptm
);
1808 localtime_r(&ti
, &tm
);
1809 strftime(date_buf
, sizeof(date_buf
),
1810 "%Y-%m-%d %H:%M:%S", &tm
);
1812 secs
= sn
->vm_clock_nsec
/ 1000000000;
1813 snprintf(clock_buf
, sizeof(clock_buf
),
1814 "%02d:%02d:%02d.%03d",
1816 (int)((secs
/ 60) % 60),
1818 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1819 snprintf(buf
, buf_size
,
1820 "%-10s%-20s%7s%20s%15s",
1821 sn
->id_str
, sn
->name
,
1822 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1830 /**************************************************************/
1833 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1834 QEMUIOVector
*qiov
, int nb_sectors
,
1835 BlockDriverCompletionFunc
*cb
, void *opaque
)
1837 BlockDriver
*drv
= bs
->drv
;
1838 BlockDriverAIOCB
*ret
;
1842 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1845 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1849 /* Update stats even though technically transfer has not happened. */
1850 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1857 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1858 QEMUIOVector
*qiov
, int nb_sectors
,
1859 BlockDriverCompletionFunc
*cb
, void *opaque
)
1861 BlockDriver
*drv
= bs
->drv
;
1862 BlockDriverAIOCB
*ret
;
1868 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1871 if (bs
->dirty_bitmap
) {
1872 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1875 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1879 /* Update stats even though technically transfer has not happened. */
1880 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1882 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
1883 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
1891 typedef struct MultiwriteCB
{
1896 BlockDriverCompletionFunc
*cb
;
1898 QEMUIOVector
*free_qiov
;
1903 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1907 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1908 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1909 if (mcb
->callbacks
[i
].free_qiov
) {
1910 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
1912 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1913 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1917 static void multiwrite_cb(void *opaque
, int ret
)
1919 MultiwriteCB
*mcb
= opaque
;
1921 if (ret
< 0 && !mcb
->error
) {
1923 multiwrite_user_cb(mcb
);
1926 mcb
->num_requests
--;
1927 if (mcb
->num_requests
== 0) {
1928 if (mcb
->error
== 0) {
1929 multiwrite_user_cb(mcb
);
1935 static int multiwrite_req_compare(const void *a
, const void *b
)
1937 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1941 * Takes a bunch of requests and tries to merge them. Returns the number of
1942 * requests that remain after merging.
1944 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1945 int num_reqs
, MultiwriteCB
*mcb
)
1949 // Sort requests by start sector
1950 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1952 // Check if adjacent requests touch the same clusters. If so, combine them,
1953 // filling up gaps with zero sectors.
1955 for (i
= 1; i
< num_reqs
; i
++) {
1957 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1959 // This handles the cases that are valid for all block drivers, namely
1960 // exactly sequential writes and overlapping writes.
1961 if (reqs
[i
].sector
<= oldreq_last
) {
1965 // The block driver may decide that it makes sense to combine requests
1966 // even if there is a gap of some sectors between them. In this case,
1967 // the gap is filled with zeros (therefore only applicable for yet
1968 // unused space in format like qcow2).
1969 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1970 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1973 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
1979 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1980 qemu_iovec_init(qiov
,
1981 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1983 // Add the first request to the merged one. If the requests are
1984 // overlapping, drop the last sectors of the first request.
1985 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1986 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1988 // We might need to add some zeros between the two requests
1989 if (reqs
[i
].sector
> oldreq_last
) {
1990 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1991 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1992 memset(buf
, 0, zero_bytes
);
1993 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1994 mcb
->callbacks
[i
].free_buf
= buf
;
1997 // Add the second request
1998 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2000 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
2001 reqs
[outidx
].qiov
= qiov
;
2003 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2006 reqs
[outidx
].sector
= reqs
[i
].sector
;
2007 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2008 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2016 * Submit multiple AIO write requests at once.
2018 * On success, the function returns 0 and all requests in the reqs array have
2019 * been submitted. In error case this function returns -1, and any of the
2020 * requests may or may not be submitted yet. In particular, this means that the
2021 * callback will be called for some of the requests, for others it won't. The
2022 * caller must check the error field of the BlockRequest to wait for the right
2023 * callbacks (if error != 0, no callback will be called).
2025 * The implementation may modify the contents of the reqs array, e.g. to merge
2026 * requests. However, the fields opaque and error are left unmodified as they
2027 * are used to signal failure for a single request to the caller.
2029 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2031 BlockDriverAIOCB
*acb
;
2035 if (num_reqs
== 0) {
2039 // Create MultiwriteCB structure
2040 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2041 mcb
->num_requests
= 0;
2042 mcb
->num_callbacks
= num_reqs
;
2044 for (i
= 0; i
< num_reqs
; i
++) {
2045 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2046 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2049 // Check for mergable requests
2050 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2052 // Run the aio requests
2053 for (i
= 0; i
< num_reqs
; i
++) {
2054 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2055 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2058 // We can only fail the whole thing if no request has been
2059 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2060 // complete and report the error in the callback.
2061 if (mcb
->num_requests
== 0) {
2062 reqs
[i
].error
= -EIO
;
2065 mcb
->num_requests
++;
2066 multiwrite_cb(mcb
, -EIO
);
2070 mcb
->num_requests
++;
2081 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2082 BlockDriverCompletionFunc
*cb
, void *opaque
)
2084 BlockDriver
*drv
= bs
->drv
;
2088 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2091 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2093 acb
->pool
->cancel(acb
);
2097 /**************************************************************/
2098 /* async block device emulation */
2100 typedef struct BlockDriverAIOCBSync
{
2101 BlockDriverAIOCB common
;
2104 /* vector translation state */
2108 } BlockDriverAIOCBSync
;
2110 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2112 BlockDriverAIOCBSync
*acb
=
2113 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2114 qemu_bh_delete(acb
->bh
);
2116 qemu_aio_release(acb
);
2119 static AIOPool bdrv_em_aio_pool
= {
2120 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2121 .cancel
= bdrv_aio_cancel_em
,
2124 static void bdrv_aio_bh_cb(void *opaque
)
2126 BlockDriverAIOCBSync
*acb
= opaque
;
2129 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2130 qemu_vfree(acb
->bounce
);
2131 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2132 qemu_bh_delete(acb
->bh
);
2134 qemu_aio_release(acb
);
2137 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2141 BlockDriverCompletionFunc
*cb
,
2146 BlockDriverAIOCBSync
*acb
;
2148 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2149 acb
->is_write
= is_write
;
2151 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2154 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2157 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2158 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2160 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2163 qemu_bh_schedule(acb
->bh
);
2165 return &acb
->common
;
2168 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2169 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2170 BlockDriverCompletionFunc
*cb
, void *opaque
)
2172 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2175 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2176 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2177 BlockDriverCompletionFunc
*cb
, void *opaque
)
2179 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2182 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2183 BlockDriverCompletionFunc
*cb
, void *opaque
)
2185 BlockDriverAIOCBSync
*acb
;
2187 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2188 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2194 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2197 qemu_bh_schedule(acb
->bh
);
2198 return &acb
->common
;
2201 /**************************************************************/
2202 /* sync block device emulation */
2204 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2206 *(int *)opaque
= ret
;
2209 #define NOT_DONE 0x7fffffff
2211 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2212 uint8_t *buf
, int nb_sectors
)
2215 BlockDriverAIOCB
*acb
;
2219 async_context_push();
2221 async_ret
= NOT_DONE
;
2222 iov
.iov_base
= (void *)buf
;
2223 iov
.iov_len
= nb_sectors
* 512;
2224 qemu_iovec_init_external(&qiov
, &iov
, 1);
2225 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2226 bdrv_rw_em_cb
, &async_ret
);
2232 while (async_ret
== NOT_DONE
) {
2238 async_context_pop();
2242 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2243 const uint8_t *buf
, int nb_sectors
)
2246 BlockDriverAIOCB
*acb
;
2250 async_context_push();
2252 async_ret
= NOT_DONE
;
2253 iov
.iov_base
= (void *)buf
;
2254 iov
.iov_len
= nb_sectors
* 512;
2255 qemu_iovec_init_external(&qiov
, &iov
, 1);
2256 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2257 bdrv_rw_em_cb
, &async_ret
);
2262 while (async_ret
== NOT_DONE
) {
2267 async_context_pop();
2271 void bdrv_init(void)
2273 module_call_init(MODULE_INIT_BLOCK
);
2276 void bdrv_init_with_whitelist(void)
2278 use_bdrv_whitelist
= 1;
2282 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2283 BlockDriverCompletionFunc
*cb
, void *opaque
)
2285 BlockDriverAIOCB
*acb
;
2287 if (pool
->free_aiocb
) {
2288 acb
= pool
->free_aiocb
;
2289 pool
->free_aiocb
= acb
->next
;
2291 acb
= qemu_mallocz(pool
->aiocb_size
);
2296 acb
->opaque
= opaque
;
2300 void qemu_aio_release(void *p
)
2302 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2303 AIOPool
*pool
= acb
->pool
;
2304 acb
->next
= pool
->free_aiocb
;
2305 pool
->free_aiocb
= acb
;
2308 /**************************************************************/
2309 /* removable device support */
2312 * Return TRUE if the media is present
2314 int bdrv_is_inserted(BlockDriverState
*bs
)
2316 BlockDriver
*drv
= bs
->drv
;
2320 if (!drv
->bdrv_is_inserted
)
2322 ret
= drv
->bdrv_is_inserted(bs
);
2327 * Return TRUE if the media changed since the last call to this
2328 * function. It is currently only used for floppy disks
2330 int bdrv_media_changed(BlockDriverState
*bs
)
2332 BlockDriver
*drv
= bs
->drv
;
2335 if (!drv
|| !drv
->bdrv_media_changed
)
2338 ret
= drv
->bdrv_media_changed(bs
);
2339 if (ret
== -ENOTSUP
)
2340 ret
= bs
->media_changed
;
2341 bs
->media_changed
= 0;
2346 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2348 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2350 BlockDriver
*drv
= bs
->drv
;
2357 if (!drv
|| !drv
->bdrv_eject
) {
2360 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2362 if (ret
== -ENOTSUP
) {
2371 int bdrv_is_locked(BlockDriverState
*bs
)
2377 * Lock or unlock the media (if it is locked, the user won't be able
2378 * to eject it manually).
2380 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2382 BlockDriver
*drv
= bs
->drv
;
2384 bs
->locked
= locked
;
2385 if (drv
&& drv
->bdrv_set_locked
) {
2386 drv
->bdrv_set_locked(bs
, locked
);
2390 /* needed for generic scsi interface */
2392 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2394 BlockDriver
*drv
= bs
->drv
;
2396 if (drv
&& drv
->bdrv_ioctl
)
2397 return drv
->bdrv_ioctl(bs
, req
, buf
);
2401 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2402 unsigned long int req
, void *buf
,
2403 BlockDriverCompletionFunc
*cb
, void *opaque
)
2405 BlockDriver
*drv
= bs
->drv
;
2407 if (drv
&& drv
->bdrv_aio_ioctl
)
2408 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2414 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2416 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2419 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2421 int64_t bitmap_size
;
2423 bs
->dirty_count
= 0;
2425 if (!bs
->dirty_bitmap
) {
2426 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2427 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2428 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2430 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2433 if (bs
->dirty_bitmap
) {
2434 qemu_free(bs
->dirty_bitmap
);
2435 bs
->dirty_bitmap
= NULL
;
2440 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2442 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2444 if (bs
->dirty_bitmap
&&
2445 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2446 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2447 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2453 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2456 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2459 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2461 return bs
->dirty_count
;